init

DESCRIPTION

@@ -0,0 +1,16 @@
+Package: lda.svi
+Title: Fit Latent Dirichlet Allocations Models using Stochastic Variational Inference
+Version: 0.0.0.9000
+Authors@R: person("Nicholas", "Erskine", email = "nicholas.erskine95@gmail.com", role = c("aut", "cre"))
+Description: Fits Latent Dirichlet Allocation models to text data efficiently using the algorithm introduced in Hoffman et al. (2011) <https://papers.nips.cc/paper/3902-online-learning-for-latent-dirichlet-allocation>.
+Depends: R (>= 3.4.4)
+License: MIT + file LICENSE
+Encoding: UTF-8
+LazyData: true
+RoxygenNote: 6.1.1
+LinkingTo: 
+    Rcpp,
+    RcppArmadillo
+Imports: 
+    Rcpp
+SystemRequirements: C++11

LICENSE

@@ -0,0 +1,2 @@
+YEAR: 2019
+COPYRIGHT HOLDER: Nicholas Erskine

LICENSE.md

@@ -0,0 +1,21 @@
+# MIT License
+
+Copyright (c) 2019 Nicholas Erskine
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

NAMESPACE

@@ -0,0 +1,7 @@
+# Generated by roxygen2: do not edit by hand
+
+export(add)
+export(lda_online)
+export(test_fn)
+importFrom(Rcpp,sourceCpp)
+useDynLib(lda.svi, .registration = TRUE)

R/RcppExports.R

@@ -0,0 +1,11 @@
+# Generated by using Rcpp::compileAttributes() -> do not edit by hand
+# Generator token: 10BE3573-1514-4C36-9D1C-5A225CD40393
+
+test <- function() {
+    invisible(.Call(`_lda_svi_test`))
+}
+
+lda_online_cpp <- function(doc_ids, terms, counts, K, passes, batchsize) {
+    .Call(`_lda_svi_lda_online_cpp`, doc_ids, terms, counts, K, passes, batchsize)
+}
+

R/add.R

@@ -0,0 +1,13 @@
+#' Add together two numbers.
+#'
+#' @param x A number.
+#' @param y A number.
+#' @return The sum of \code{x} and \code{y}.
+#' @examples
+#' add(1, 1)
+#' add(10, 1)
+#' @export
+add <- function(x, y) {
+  x + y
+}
+

R/lda.svi-package.R

@@ -0,0 +1,10 @@
+#' @keywords internal
+"_PACKAGE"
+
+# The following block is used by usethis to automatically manage
+# roxygen namespace tags. Modify with care!
+## usethis namespace: start
+#' @useDynLib lda.svi, .registration = TRUE
+#' @importFrom Rcpp sourceCpp
+## usethis namespace: end
+NULL

R/lda_online.R

@@ -0,0 +1,40 @@
+#' Fit a Latent Dirichlet Allocation to text data 
+#'
+#' @param dtm A data frame with three columns. The first column is a vector of document ids (a character vector), the second is a a vector of terms (a character vector), and the third is a vector of counts (an integer vector).
+#' @param passes How many times we look at each document
+#' @param batchsize The size of the minibatches
+#' @param eta hyperparameter
+#' @param alpha hyperparameter
+#' @param kappa learning rate parameter
+#' @param tau_0 learning rate parameter
+#' @param K The number of topics
+#' @export
+lda_online <- function(dtm,passes=1,batchsize=256,K,eta=1,alpha=1,kappa=0.7,tau_0=1024){
+
+	docs <- dplyr::pull(dtm,document)
+	terms <- dplyr::pull(dtm,term)
+	counts <- dplyr::pull(dtm,n)
+
+	doc_ids <- seq(0,length(unique(docs)))
+	names(doc_ids) <- unique(docs)
+
+	term_ids <- seq(0,length(unique(terms)))
+	names(term_ids) <- unique(terms)
+
+	cat('launching cpp code')
+	res_list <- lda_online_cpp(doc_ids[docs],term_ids[terms],counts,K,passes,batchsize)
+
+	gamma <- res_list$Gamma
+	lambda <- res_list$Lambda
+
+	colnames(gamma) <- seq(1:ncol(gamma))#topic labels
+	rownames(gamma) <- unique(docs)
+
+	colnames(lambda) <- unique(terms)
+	rownames(lambda) <- seq(1:nrow(lambda))
+
+
+
+	res_list#TODO: tidy output 
+
+}

R/test.R

@@ -0,0 +1,8 @@
+#' @export
+test_fn <- function(dtm){
+
+  #print(class(dtm))
+  #print(dim(dtm))
+
+  print(dtm[[1]])
+}

README.md

@@ -0,0 +1,29 @@
+## This is very experimental for now 
+
+# lda.svi
+
+This R package fits latent dirichlet allocation models to data using the stochastic variational inference method introduced in [this paper](https://papers.nips.cc/paper/3902-online-learning-for-latent-dirichlet-allocation) by Matthew Hoffman and coauthors. This method allows LDA models to be fit considerably faster, and using considerably less memory, than with the previous batch variational Bayes method. As far as I can tell, this is the only R package implementing this method. The key functions are implemented in C++ for speed.
+
+## Non-R Dependencies
+
+* A C++ compiler with a reasonably modern version of the standard library.
+
+## Philosophy
+
+The interface is designed with Hadley Wickham's [tidy data principles](https://vita.had.co.nz/papers/tidy-data.pdf) in mind, and therefore fits in nicely with the [tidytext](https://github.com/juliasilge/tidytext) package by Julia Silge and David Robinson, which I recommend for preprocessing text and postprocessing model output. 
+
+## Installation
+
+I intend to submit this to the CRAN soon, but in the meantime you can install it by running
+
+```{r}
+#install.packages('devtools')
+devtools::install_github("nerskin/lda.svi")
+```
+
+## TODO:
+
+* Check that the output makes sense 
+* Tidy the output
+* Add documentation/vignette
+* Submit to CRAN

build.R

@@ -0,0 +1,5 @@
+#!/usr/bin/env Rscript
+
+Rcpp::compileAttributes()
+pkgload::load_all()
+roxygen2::roxygenise()

build_notes.txt

@@ -0,0 +1,3 @@
+* ~~~Make sure to use roxygen2 version 6.0.1 - later versions give mysterious errors. See https://github.com/klutometis/roxygen/issues/822~~~
+
+Weird errors seem to be fixed by always running devtools::install() before roxygenise()

build_test.bash

@@ -0,0 +1,5 @@
+Rscript -e "Rcpp::compileAttributes()"
+
+R CMD INSTALL .
+
+Rscript -e "lda.svi:::lda_online(rep(0:1,500),rep(0:1,500),rep(10,1000),2)"

lda.svi.Rproj

@@ -0,0 +1,17 @@
+Version: 1.0
+
+RestoreWorkspace: Default
+SaveWorkspace: Default
+AlwaysSaveHistory: Default
+
+EnableCodeIndexing: Yes
+UseSpacesForTab: Yes
+NumSpacesForTab: 2
+Encoding: UTF-8
+
+RnwWeave: Sweave
+LaTeX: pdfLaTeX
+
+BuildType: Package
+PackageUseDevtools: Yes
+PackageInstallArgs: --no-multiarch --with-keep.source

src/.gitignore

@@ -0,0 +1,3 @@
+*.o
+*.so
+*.dll